I'm noticing recall rate drops when filtering by payload index. I guess that makes sense since the effective m value for the HNSW graph drops as well. This can be some what mitigated by a larger ef search. I'm wondering if you guys have thought about using the query planner to estimate the effective m value for the query and adjusting the ef search value accordingly?

i'm facing cuda error , while trying to finetune mpnet all v2 with 1Lakh rows data, 4gb gpu

Hi @everyone I have recently been working more with Qdrant on a couple of projects and the results have been pretty amazing. Kudos to the team on the features and seamless integration. One issue I'm currently facing though is difficulty at accessing Remote qdrant container. I deployed the latest version of qdrant to AWS using aws lightsail, I'm having issues uploading payload and vectors to the cloud-hosted container using python qdrant client. My guess is CORS but I'm not too sure if that the is the cause. Any help on this will highly appreciated Thanks

For example, using this filter: 'filter': {'must_not':[{'key':'type', 'match':{'value':'quote'}}], 'must':[{'has_id': [14, 30,69, 1405, 560, 900]}] -- seems to ignore the must_not aspect of the filter, and returns results from the ids, regardless of whether their value is quote or not

@mention I'm also open to storing the actual text as payload in qdrant, with each vector. Will indexing this payload allow me to query the text like a full-text search

Hi everyone! We are publishing a curated list of awesome practical metric learning tools, libraries, and materials - https://github.com/qdrant/awesome-metric-learning There are a lot of academic research papers around the topic but much fewer practical guides and tutorials. So we decided to help adopt metric learning by putting together references to all relevant materials. If you know an exciting article, tool, or blog post which helped you apply metric learning - feel free to PR your proposal!

Hi everyone! I have a good proposition about RAM and CPU analysis. I tried to integrate tracy profiler tool (https://github.com/wolfpld/tracy) and it works perfectly. I request to make a presentation to core team. I want to show this tool, tell about functionality, advantages and how to integrate it.

conversation-10

conversation-0